【作者】 贾文峥；

【导师】 毛保华； 何天健；

【作者基本信息】 北京交通大学 ， 系统分析与集成， 2010， 博士

【摘要】 衔接多个方向的大型客运站接发车作业密集,咽喉区交叉干扰多,通常构成行车组织的瓶颈,例如广州站、武汉站、郑州站、北京站及成都站等。这类大型客运站的能力既要能满足运行图中规定的接发车作业,又要实现一定的服务水平。在相同的时空范围内,随着接发车作业数量的增加,车站对列车作业的服务水平呈下降趋势；当车站能力不能满足运行图的需求时,一些列车就会产生到达或出发晚点。在车站平面图和列车运行图已知,以及车站能力满足图定列车作业的前提下,考虑到列车运行过程中的扰动,不同的接发车方案可能产生不同的效果,例如一个方案的抗干扰能力强于另外一个方案。在车站能力富余的前提下,研究如何获取抗干扰能力较好的接发车方案值得探讨,这将为分析车站能力与运行图抗干扰性的关系提供参考。因此,为获取高质量的列车接发车方案,需要研究车站接发车作业的进路分配问题。铁路客运站的进路分配问题是编制旅客列车运行计划的一部分,目的是为运行图中的列车分配无冲突的进路和站台,同时满足车站作业和运输组织的要求。进路分配问题中包含由列车运行图和车站平面图引起的时间和空间约束,求解进路分配问题的一个基本要求是疏解列车作业之间的时空冲突,同时考虑车站设备利用以及列车服务水平,进路分配问题的解称为进路分配方案。既有方法中的目标函数包括列车等级总权重最大化、到发线运用效用最大化、分配进路总权重最大化以及列车晚点时间最小化等,部分研究提到进路分配方案与列车作业抗干扰性的关系问题,但未能充分考虑车站能力富余条件下以优化抗干扰性为目标的进路分配问题。本文采用分层方法探讨进路分配问题与列车作业缓冲时间的关系,取得较好效果。首先以列车运行计算数据为基础,考虑列车接发车作业占用进路的详细时间,建立进路分配问题的约束满足模型。然后提出评价进路分配方案的定量指标,并利用仿真方法分析进路分配方案的抗干扰性,最后提出根据评价指标对进路分配方案进行局部优化的可行性。主要研究内容和结论包括以下五个方面。(1)构造列车作业时间窗重叠图模型。利用列车运行计算获取列车接发车作业占用轨道区段的时间窗信息,通过判断列车作业时间窗之间的关系,构造列车作业时间窗重叠图。利用该模型可以将列车作业按照时间窗重叠关系分为若干个簇,在求解进路分配问题时以这些簇为单位疏解列车作业的空间冲突。因此,列车作业时间窗重叠图模型可以将大规模的进路分配问题进行分解。(2)建立进路分配问题的约束满足模型,并利用求解约束规划(ConstraintProgramming, CP)问题的启发式搜索方法疏解列车作业的空间冲突。约束条件分为硬约束和软约束两类,硬约束表示不可违反的约束条件,包括股道作业间隔时间约束和进路冲突约束；软约束表示偏好,包括上下行约束、列车等级引起的约束、列车接续引起的约束、动车组运用计划引起的约束。求解过程分为约束识别、值排序以及回溯搜索(Backtracking)三个步骤。(3)建立评价进路分配方案的定量指标体系。指标主要包括“使用站台的偏好”、“缓冲时间数”、“最小缓冲时间”、“车站设备利用率”以及瓶颈和晚点传播等内容。由于列车停留的股道表明了旅客乘降及技术作业的可用站台,因此“列车分配最优股道的总权重”反映了车站为列车提供的服务水平。两项作业之间不存在缓冲时间表示它们是完全的平行作业,存在缓冲时间则表示相反情况,即前一项作业受到的扰动大于缓冲时间时,后一项作业就会被延误,因此“缓冲时间数,,表示进路分配方案中存在的潜在冲突数。“最小缓冲时间”是指进路分配方案中所有缓冲时间的最小值,它表示方案中最脆弱的两项作业,也表示整个方案能容忍的不引起晚点传播的最大扰动。“车站设备利用率”反映了股道与咽喉区轨道区段的利用率,根据这些指标可以发现车站平面图中对接发车作业构成瓶颈的道岔组或轨道区段。(4)利用有色时间Petri网技术建立进路分配方案的动态模型,并进行扰动分析。接发车作业的基本单元是列车占用和释放轨道区段的活动,在构造列车活动模型的基础上,建立单列车作业模型、两列车作业模型以及大规模进路分配方案的分层Petri网模型。模型可以描述列车运行图、车站轨道区段与进路以及进路分配方案,列车作业缓冲时间被隐式包含在模型中。通过对模型中的列车活动施加扰动,并收集模型中的列车实际到发时间信息,就可以获取扰动造成的晚点传播,从而识别方案中的瓶颈,并利用列车晚点时间等信息比较方案的抗干扰性。(5)基于实际车站平面图的案例分析表明缓冲时间的分布及大小是影响进路分配方案抗干扰性的根源,这一发现为进路分配方案的局部最优化及制定调车作业方案提供数据支撑。以包含1045条进路的车站平面图和持续时间为1小时的列车运行图为案例,利用既有的WNPP方法和本文的约束规划方法求解相同规模的问题,两种方法所得方案的最小缓冲时间分别为17秒和188秒。分别对两个方案施加360秒的扰动,约束规划方法所得方案中列车的平均出发晚点时间比WNPP方法所得方案中的列车晚点少35.07%,因此约束规划方法能得到抗干扰性较好的方案。这进一步验证了约束规划方法求解进路分配问题的有效性。更多还原

【Abstract】 The large-scale passenger rail stations are normally the bottlenecks of executing the scheduled train services, such as the Guangzhou station, Wuhan Station Zhengzhou station, Beijing station and Chengdu station. The infrastructure needed by the scheduled arrival and departure events must be available and some level of service should be achieved. The level of service tends to decrease when the scheduled trains increase and the arrival/departure delays occur when station capacity can not satisfy the trains demand. However, the delay-tolerances for two routing schemes might be different when there is surplus station capacity. Therefore, close attention should be paid to achieving delay-tolerable routing schemes.Routing trains through rail stations is a sub-problem of scheduling passenger trains, and the aim of the Route Allocation Problem (RAP) is to allocate platforms and conflict-free inbound/outbound routes to trains while satisfying the specific train operation guidelines and ensure the station operation safety. The RAP is imposed on many temporal and spatial constraints which are caused by train timetable and station layout. The existing objectives include maximizing the total weight of train grade, maximizing the effectiveness of the platform tracks, maximizing the total weight of routes and minimizing the total delays. Some literatures mention that there is an internal delay-tolerable ability for each Route Allocation Solution (RAS). However, the problem of finding delay-tolerable RAS is rarely investigated. So, the overall objective of the RAP is to find one or some robust RAS taking consider of stochastic disturbances to trains in real-life operations.A hierarchical approach is adopted to tackle the above RAP and the expected results are obtained. Firstly, the constraints satisfying model of the RAP is built and the blocking times of the arrival and departure events are embedded in the model by train movement calculation. Secondly, a group of quantitative indices are put forward to evaluate the obtained RAS from the above model and a RAS from other approaches can also be evaluated according to the proposed indices. Then, a simulation models is constructed to analyze the delay-tolerance of the above solutions. Finally, the feasibility of locally optimizing a RAS is discussed based on the evaluation of the RAS. The main contents of the thesis are summarized as follows.(1) The time-windows overlapping model for arrival and departure events of trains is constructed. The blocking times of the events are achieved by the train movement calculation and the relation of the time-windows for two events can be made clear. The arrival and departure events of trains can be divided into some cliques and the essence of the RAP is to resolve the routes conflicts between any two events in a clique. Therefore, the RAP can be decomposed into small parts by means of the here proposed model.(2) The constraints satisfying model of the RAP is constructed and the heuristic searching for constraints programming (CP) is exploited to resolve the temporal conflicts among arrival and departure events. The constraints are divided into hard and soft ones. The former states the minimum interval between two events accessing the same platform-track and the constraints on two events accessing the same track section in throat areas, while the latter indicates the preference, including the constraints form the up-down rule, trains grade, trains connection and locomotives plans. Three steps, detecting constraints, ordering values and backtracking are followed to resolve the model.(3) A group of quantitative indices is proposed to evaluate the obtained RAS. The indices cover the platform-tracks preference, the number of buffer times, the minimum buffer time and the infrastructure utilization. The platform-tracks preference indicates the level of service of trains. Two events are totally parallel if there is no buffer time between them, while the buffer time means the maximal perturbation which the former event can bear and will not cause delay propagation, therefore the number of buffer times is the number of potential conflicts in a RAS. The minimum buffer time means the weakest part of a RAS. The bottleneck area in the station layout can be discovered by means of the infrastructure utilization.(4) The dynamic model of RAS is developed by the colored timed Petri Nets (CTPN) technology to simulate dynamic behavior of RAS. The train activity model, the single and double train model, and the hierarchical model for large-scale station layout are established. The buffer times are embedded in the model, and the bottlenecks in a RAS can be detected and some indices about delay propagation are also achieved by performing disturbances analysis.(5) The cases study indicates that the buffer times is the root determining the delay-tolerance of a RAS and the finding provide data support to optimize a RAS locally and to make shunting plans. The cases based on a practical station layout with 1045 routes and a one-hour timetable show that the minimum buffer times in two RAS obtained by WNPP and CP are 17 seconds and 188 seconds. The average departure delays for the RAS obtained by CP is 35.07% less than that for the RAS achieved by WNPP when they suffer an average disturbance of 360 seconds. So, CP is an effective approach to resolve the RAP and can obtain relatively robust solutions.更多还原